Anthropic is thrilled to present Claude 2, its latest model designed to enhance performance, safety, and functionality. Claude 2 is your friendly and enthusiastic AI assistant, ready to assist you with various tasks using natural language instructions. With improved capabilities, longer responses, and accessible through our API and beta website, claude.ai, Claude 2 aims to provide a seamless and productive user experience.
Listening to user feedback, they have incorporated valuable insights into the development of Claude 2. Users have praised its conversational abilities, clear explanations, reduced harmful outputs, and expanded memory capacity. Claude 2 now retains more information, enabling comprehensive and insightful responses.
Notably, Claude 2 excels in coding, math, and reasoning. It achieved an impressive score of 76.5% in the multiple-choice section of the Bar exam, surpassing its predecessor, Claude 1.3. When compared to college students applying to graduate school, Claude 2 scores above the 90th percentile on the GRE reading and writing exams and performs similarly to the median applicant on quantitative reasoning.
Claude 2 now supports larger input and output lengths, allowing users to input up to 100,000 tokens per prompt. This enables processing of extensive technical documentation, books, and other lengthy texts. Additionally, Claude 2 can generate longer documents, including memos, letters, and stories, in one go.
Safety remains a top priority, and Anthropic has taken significant measures to ensure Claude 2 produces harmless responses. Through internal evaluations and rigorous testing, Claude 2 demonstrates a twofold improvement compared to Claude 1.3 in generating safe outputs.
Claude 2 powers the chat experience and is currently available in the US and UK. the company is actively working towards global availability. Users can create an account and engage in natural language conversations, leveraging Claude 2's capabilities for various tasks. It encourages users to explore and experiment while referring to the provided tips for optimal interaction.
Anthropic is proud to collaborate with businesses already utilizing the Claude API. Jasper, a generative AI platform, recognizes Claude 2's effectiveness in diverse use cases, particularly for long-form low latency applications. Sourcegraph, a code AI platform, leverages Claude 2's improved reasoning abilities for their coding assistant, Cody, providing developers with accurate answers and valuable insights.
The company seeks feedback from its users and considers them valuable as they responsibly deploy their products. While their chat experience is in open beta, they acknowledge that AI models, including Claude, may occasionally generate inappropriate responses. AI assistants excel in everyday situations, such as summarizing and organizing information. However, they should not be relied upon for situations involving physical or mental health and well-being.
With Claude 2, Anthropic aims to provide an AI assistant that is not only highly capable but also friendly, trustworthy, and dedicated to assisting you effectively. Anthropic expects the support and feedback of its users as it continues to enhance Claude's capabilities and safety features.
Data Phoenix Newsletter
Join the newsletter to receive the latest updates in your inbox.